Diagnosis result fusion algorithm in fault diagnosis system using improved D-S theory

Abstract

Adoption of Dempster-Shafer (D-S) evidence theory in multi-agent system (MAS) fault diagnosis is an effective way to improve diagnosis performance by fusing multiple diagnosis results. To solve the one-vote veto problem and improve robustness in D-S evidence theory, weight for diagnosis agent is introduced. A dynamic weight adjustment mechanism based on diagnosis experience accumulation is put forward to ensure that the diagnosis agents with respective diagnosis accuracy proportionally contribute to the final fusion diagnosis result. Experiment results show that the algorithm possesses high effectiveness, reliability and astringency in fault diagnosis.